Login Close Login Home
Login mdi-account-circle

MKG: Medical Knowledge Graph is a medical ontology built on ICD-10

To conduct a medical examination by software it is necessary to:

mdi-chevron-right or reproduce past exchanges between a doctor and his patient (Machine Learning),

mdi-chevron-right either follow an order of questions determined in advance (decision tree)

mdi-chevron-right either conduct a reasoning according to a given strategy applied to knowledge (expert system)

Because medicine is constantly evolving and we wanted maximum scalability, we decided to attack the everest by the north face :-) and we chose to work on the structuring of medical knowledge, creating our own medical ontology describing the types of relationships between diseases, syndrome, risk factor, drugs, complementary examinations, symptoms, questions, answers, surgical procedures.

The MKG now contains 92,000 of these entities, each linked to a recognized classification or codification where it exists (CIM-10, DSM-5, Orphanet, DRC, ATC/ANSM...)

Towards a self-learning medicine treatise?

The MKG, made up of reliable databases verified by the scientific community (CIM-10, Orphanet, ATC, DSM-5, DRC...) is our repository facilitating interoperability between any solution of the ANAMNESE platform and the solutions or DPI of other medical software.

Enriched since 2017 with the help of medical experts (interns, doctors, etc.), the Medical Knowledge Graph contains to date

mdi-chevron-right 21,000 diseases or syndromes referenced in WHO ICD-10 (in its 2020 version of ATIH)

mdi-chevron-right 5000 symptoms (by location, intensity, temporality...)

mdi-chevron-right 6000 risk factors

mdi-chevron-right 19000 questions and answers

mdi-chevron-right 16000 ATC drug codes

mdi-chevron-right ...

Each use of Anamnesis (patient responses to questions - diagnosis chosen by the doctor) helps to refine the prevalences (and conditional probabilities) of a disease knowing that the patient has such symptoms and risk factors at such time of year in a particular region...

By using Anamnesis, you help us build a knowledge base that could eventually become a digital medicine treatise, constantly kept up to date with real-life data.

the MKG allows Explicable Artificial Intelligence

Artificial Communications (machine learning such as machine learning) allows for great scientific advances, but also has great flaws:

mdi-chevron-right it reproduces the biases of the samples on which she learned

mdi-chevron-right unlearning him is complicated

mdi-chevron-right its results are hardly explainable

By relying on the modelling of knowledge comprehensible by medical experts, and using reasoning mechanisms on this knowledge, Anamnesis relies primarily on techniques of Symbolic Artificial Intelligence (type expert system) that make its results explicable to the scientific community.

But as it would be a shame to ignore real life data, we are currently working to combine these two technologies to offer a HybridArtificial Intelligence, combining logical reasoning with machine learning, much like a doctor complements his theoretical knowledge of medicine read in the articles, with his practical experience with patients.

Interoperability through the use of repository

Whether it's drugs, surgeries, diseases, risk factors, symptoms; any medical entity within ANAMNESE is codified, using internationally recognized references as much as possible.

This allows:

mdi-chevron-right to structure information

mdi-chevron-right to facilitate exchanges with third-party systems

mdi-chevron-right to facilitate the use of data in real life, and thus to facilitate the search by limiting cleaning operations (the imposed stage of big data).

Dynamic and personalized medical questionnaire

Thanks to the MKG, it is possible to create as many specialized questionnaires as there are medical situations (emergency, screening, regulation, etc.).

More info on our multilingual patient questionnaires

Differential diagnosis

Symptoms that don't usually express themselves together?

Doubt about a rare disease?

The MKG can also be used to produce a differential diagnosis, and for a set of symptoms, and a type of patient, you suggest the most likely diagnoses, and for each, the symptoms or risk factors on which to question the patient

Do you have a plan?

Contact us

You want to learn about the other components of the e-health platform

Back to the Platform's components
Who are we? News Frequently Asked Questions Contact us Confidentiality policy Terms and conditions mdi-facebook mdi-linkedin-box mdi-twitter mdi-youtube 2020 - © Anamnèse